Improving patcher performance (and avoiding slab to matrix for alpha blending)
Hi all - I've got a patcher that's nearly done, but I'm not happy with the performance. I read about the vade optimizations (http://abstrakt.vade.info/?p=147) - and watched the cycling74 tutorial on optimization - http://www.youtube.com/watch?v=8xwdbdljAKA. I've tried various combinations of things to get my framerate up - which you can see in this patch, such as qlim on the preview monitor, unique 1 on the movies, disabling onscreen on the preview window, etc.
One thing I have a feeling is that, while I'm rendering a syphon live camera feed and a movie together on a slab, I later use a matrix to alpha blend that with another video. I feel like there's a way to accomplish this while maintaining the use of slabs, but I haven't figured it out.
At any rate, if anyone is willing to have a peak at the patch and suggest some optimizations, or even point me in the right direction, I'd be incredibly grateful. Thanks!
The patch doesn't have any video or live feed so it's impossible to test.
Yes - the live videos are pulled in from a text file - each one that I'm using is about 2GB so I can't attach them. I was more just hoping someone could look at the patch and see if there's anything I'm doing that seems like it would cause a performance bottleneck, or could be optimized.
you will definitely get better performance if you remove the readback to jit.matix.
alphablending is very simple to achieve in GL land (once you know a few details).
simply overlay two gl.videoplanes, with @depth_enable 0, @blend_enable 1, and control the layer order with @layer attribute.
This is great! Definitely faster. Is there any way to still have a preview of the entire composition now that things are separated in to two video planes? Before it was stemming off the combined matrix - but now that combination is only happening from a blend parameter off two different objects.